Controle de qualidade de biodieseis de soja e pinhão manso usando espectroscopia no infravermelho médio e análise discriminante por quadrados mínimos parciais

Abstract

Biodiesel has been highlighted for being an alternative source of energy since it can replace or be added to diesel oil. This renewable fuel is already consolidated in Brazil being of interest to the Government of Mozambique to acquire production technologies and quality control throughout the process to distribution phase. In Brazil, supervision is done by the Brazilian National Agency of Petroleum, Natural Gas and Biofuels (ANP) that establishes quality parameters. The parameters of ANP are based on standards of the signatory institutions of the International Metrology System, for example, the Standards of American Standard for Testing Materials. Among the various analytical techniques for quality control of biodiesel stands out mid-infrared spectroscopy because it is fast, non-destructive analysis, can be applied in situ, at all stages of production and inspection. However, the potentiality of MIR spectroscopy become unlimited when combined with chemometric methods to extract relevant information of complex data, making them simpler to be applied in laboratories routine. Starting from the above assumptions, in this work, MIR spectroscopy technique allied to partial least square discriminant analysis (PLS-DA) was applied, aiming to develop methodologies able to classify samples of biodiesels according its origin the biodiesels methylic and ethylic, pure B100 and mixtures with diesel in proportion of 5% (v/v) mixture known as B5, of soybean oil, used frying oil, jatropha oil. In addition, classify samples of pure B5, and adulterated with gasoline, residual automotive lubricant oil, soybean oil and used frying oil. The models were built using PLS1-DA and PLS2-DA, the second being to demonstrate the advantage of simultaneous classification of different types of biodiesels in the same model, which reduces the analysis time, important factor to take decision in production line and laboratories. All models developed showed high sensitivity and specificity with values equal 1 (maximum classification), meaning 100% correct classification in all sample of the test set. Thus, the methodologies proposed can be applied in quality control of these biofuels and their blends with diesel.